The rights of artificial intelligence systems have taken a step backwards with the recent Full Court of Australia’s decision to reverse Thaler v Commissioner of Patents [2021] FCA 879. Australia is now following the rest of the world in requiring patents to have a human inventor.
In 2019, Dr. Stephen Thaler filed patent applications in various jurisdictions including Australia, South Africa, the European Patent Office, United Kingdom, United States and New Zealand. So far so good. What was unusual was that the patent applications listed Dr. Thaler’s artificial intelligence ‘DABUS’ as the inventor.
Like the rest of the world (with the technical exception of South Africa), the Deputy Commissioner rejected the application in the first instance on the basis that the inventor must be a human.
Unlike the rest of the world, when Dr. Thaler appealed the Deputy Commissioner’s decision, the Federal Court agreed that AI program DABUS could be the inventor, considering that (amongst other things):
- it was a fallacy that an inventor could only be human;
- the Patents Act 1900 (Cth) (Patents Act) did not specifically reject the possibility of non-human inventors;
- the Deputy Commissioner had failed to consider the evolving nature of patentable inventions and their creators; and
- if the appeal was not upheld, then an invention that had been developed by an artificial intelligence could never be protected.
On appeal, the Full Federal Court has found that DABUS cannot be listed as the named inventor for the patent application because:
- the sub-sections within section 15 of the Patents Act, which cover who may be granted a patent, each contemplate the inventor either being a person or having the right to pass title to the patent applicant. Only an entity with legal identity can assign or grant rights and an artificial intelligence is not a legal entity. Given this, an inventor must be a natural person;
- the term ‘inventor’ has its ordinary meaning, being a natural person, and should have the same meaning throughout all of section 15 (and the Patents Act generally). There is no logic in ascribing a different meaning to ‘inventor’ to the various sub-sections of section 15; and
- this matched the policy objectives of the Patents Act and the historical use of patents, including the goal of granting a reward for the development of an invention.
The Full Federal Court did not accept that its decision meant that applications to register an invention ‘developed’ by an artificial intelligence would necessarily fail for lack of an ‘inventor’. Whether an invention has an ‘inventor’ needs to be considered at law in each instance.
As such, it remains an open question as to whether, if such an invention is capable of registration as a patent, the inventor would be the person who trained the artificial intelligence, the person who decided upon the specific inputs to give rise to the invention, or someone else entirely.
For advice regarding protection of inventions and patent strategy generally, please contact Hamilton Locke’s intellectual property team.
